Individual variations in ‘brain age’ relate to early-life factors more than to longitudinal brain change
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
This paper is of interest to scientists within the field of lifespan developmental neuroscience. The data analysis is rigorous and the conclusions are justified by the data. The key claims of the manuscript are directly related to, and support, a more reasonable interpretation of previous known findings.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 agreed to share their name with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Neuroscience (eLife)
Abstract
Brain age is a widely used index for quantifying individuals’ brain health as deviation from a normative brain aging trajectory. Higher-than-expected brain age is thought partially to reflect above-average rate of brain aging. Here, we explicitly tested this assumption in two independent large test datasets (UK Biobank [main] and Lifebrain [replication]; longitudinal observations ≈ 2750 and 4200) by assessing the relationship between cross-sectional and longitudinal estimates of brain age. Brain age models were estimated in two different training datasets (n ≈ 38,000 [main] and 1800 individuals [replication]) based on brain structural features. The results showed no association between cross-sectional brain age and the rate of brain change measured longitudinally. Rather, brain age in adulthood was associated with the congenital factors of birth weight and polygenic scores of brain age, assumed to reflect a constant, lifelong influence on brain structure from early life. The results call for nuanced interpretations of cross-sectional indices of the aging brain and question their validity as markers of ongoing within-person changes of the aging brain. Longitudinal imaging data should be preferred whenever the goal is to understand individual change trajectories of brain and cognition in aging.
Article activity feed
-
-
Evaluation Summary:
This paper is of interest to scientists within the field of lifespan developmental neuroscience. The data analysis is rigorous and the conclusions are justified by the data. The key claims of the manuscript are directly related to, and support, a more reasonable interpretation of previous known findings.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 agreed to share their name with the authors.)
-
Reviewer #1 (Public Review):
Because researchers who use "BrainAge" (the prediction error between one's chronological age and an estimated brain health age) as their metric of choice must rely on the assumption that this estimated BrainAge index is a reflection of accumulated within-person effects of time on that person, this assumption needs to be verified or falsified. The current study aimed to test that assumption by utilizing several large-scale datasets that contained both cross-sectional and longitudinal data collected from structural MRI scanning. BrainAge studies by definition (because they are based on comparing an individual to a sample "norm") utilize cross-sectional data. However, cross-sectional estimates of age-related brain differences has been shown in other studies to not be a reliable predictor of true within-person …
Reviewer #1 (Public Review):
Because researchers who use "BrainAge" (the prediction error between one's chronological age and an estimated brain health age) as their metric of choice must rely on the assumption that this estimated BrainAge index is a reflection of accumulated within-person effects of time on that person, this assumption needs to be verified or falsified. The current study aimed to test that assumption by utilizing several large-scale datasets that contained both cross-sectional and longitudinal data collected from structural MRI scanning. BrainAge studies by definition (because they are based on comparing an individual to a sample "norm") utilize cross-sectional data. However, cross-sectional estimates of age-related brain differences has been shown in other studies to not be a reliable predictor of true within-person change over time.
Applying several instantiations of machine learning to a training set (n = 38,582) and a test set (n = 1372) of data from the UK Biobank database the authors compared cross-sectional BrainAge to true longitudinal change in brain structural features and computed a longitudinal BrainAge to compare to true longitudinal change in brain structure. This was then applied to an independent replication dataset (Lifebrain n = 3292).
Importantly, the study results found that BrainAge score was not predictive to true brain aging, neither cross-sectionally nor longitudinally. The results did show that BrainAge was significantly predicted by measures available by the time of an individual's birth (birth weight n =770 and a polygenic score taken from a GWAS n = 38,163). These two findings taken together suggests that BrainAge index is not a reliable indicator of brain aging (over time) and certainly does not reflect "accelerated aging", but rather BrainAge index appears to be influenced by factors already present at birth.
Given the mis- and over-interpretation of BrainAge scores, and the mis-representation of what cross-sectional designs can say about true longitudinal change that is rampant in the field, this study is timely and important. As such, it has many strengths including a very large sample size across several independent datasets, a good and proper use of computation of BrainAge scores (including correcting for age-bias and the inclusion spread), use of linear and GAM fits, and generalizing results across other machine learning algorithms and across independent datasets.
On the whole, this is a successful study and the authors achieved their aims. While the interpretation and discussion of the findings does not overstep, some terminology could be clearer or more accurately defined. For example, using the term "early life influences" seemed a bit over-encompassing being only indexed by weight at birth and a genetics score. The reason this terminology precision is important is because early life influences connotes a rich set of biological and environmental variables during childhood and adolescence that impact one's ultimate brain health in middle and older age. I recommend changing to a term that better reflects "indices already available at birth". Otherwise this oversteps, conceptually. Another weakness is that the sheer amount of detail and variables involved in the study are not provided until very late in the manuscript (since the format of Intro-Results-Methods-Discussion-Supplemental is used). This manuscript was incredibly difficult to parse, and this is from a reviewer directly in this area of study -- a reader not in this direct field has little chance of successfully parsing this manuscript as currently written. I found it to be a necessity to simultaneously read the Methods and Supplemental Info to be able to read and understand even the abstract, intro, and discussion. There were minor confusions and errors in the figures and captions, etc that I detail in the review to authors.
Having said that, other than readability, the strengths far outweigh the weaknesses of this submission and it stands to have a high influence on the research in this field, aimed particularly at urging caution in users of (and readers of) the BrainAge metric, and as a higher, more generalized point, the limitations of assumptions of true aging and brain change from cross-sectional data.
-
Reviewer #2 (Public Review):
This work provides a new set of evidence on the necessity of longitudinal data for experimental designs to understand individual changes of brain aging. Such evidence were derived with a large-scale dataset UKB and replicated in another, and thus offered strong statistical power to detect small or moderate effects. To achieve their aims, the authors employed the BrainAge method and demonstrated the lack of its association between cross-sectional and longitudinal derivations. Adults' BrainAge metrics were related to the measurements of their birth weight and polygenic scores. This calls cautions in interpreting cross-sectional indices of the aging brain as well as concluding their validity as markers of individual-level brain aging process.
-
-